A radio frequency (RF) signal transmission facility such as a base station or a mobile station in wireless communication employs a digital pre-distorter (DPD) that executes a process of correcting signal distortion at a digital level (see, e.g., Japanese Laid-Open Patent Publication No. 2006-295992). Recently wideband communication such as that employing wideband code division multiple access (W-CDMA) or orthogonal frequency division multiplexing (OFDM) has been become prevalent.
With the spread of wideband communication, the correction of memory effect of distortion characteristics of a power amplifier (PA) is desirable. Types of general-purpose distortion models that express the distortion characteristics of the power amplifier include a type of model that combines expressions of the memory effect. For example, assuming that a time interval is represented by “Δ”, a general-purpose distortion model that expresses a distortion characteristic of a power amplifier is expressed as equation (1) below using a memory polynomial expression (the quadratic term of the Volterra series).
                                          v            0                    ⁡                      (            t            )                          =                              ∑            k                    ⁢                                    ∑              n                        ⁢                                          a                                  k                  ,                  n                                            ⁢                                                v                                      i                    ⁢                                                                                  ⁢                    n                                    n                                ⁡                                  (                                      t                    -                                          k                      ⁢                                                                                          ⁢                      Δ                                                        )                                                                                        (        1        )            
In equation (1), “vo(t)” represents distortion of a signal amplified by the power amplifier. “t” represents time. “ak, n” represents a coefficient for each amplifier-related factor causing distortion of the signal. “vin” represents the signal before amplification by the power amplifier. “t−kΔ” represents the memory effect of the power amplifier.
However, in the conventional technique, use of a higher-order series is desired in the model expressing the distortion characteristics of the amplifier. Therefore, a tremendous number of combinations of memory effects and distortion properties must be calculated to generate a correction signal, arising in a problem of a tremendous amount of calculation. Especially, in a modulation scheme whose peak power is large relative to its average power such as W-CDMA or OFDM, the memory effect of the distortion characteristics is significant and the amount of the calculation becomes even more significant. Therefore, calculation of the distortion characteristics of the amplifier consumes a great deal of time and the convergence of the correction signal is degraded. Power consumption also becomes tremendous because the amount of calculation becomes tremendous.
If the order of the model is reduced to reduce the amount of calculation, the memory effect of the amplifier is not sufficiently reflected on the correction coefficient. Therefore, a problem arises in that the signal distortion caused by an amplifier that has a memory effect can not be corrected accurately.